Another update - I've switched from CUDA 6.5 to 5.5, however the error remains the same:
Error: unspecified launch failure launching kernel kClearForces
cudaFree GpuBuffer::Deallocate failed unspecified launch failure
Is it possible that because both my cartesian restraints (only on chain A) AND NMR restraints (tethering chain B to chain A) affect some of the same residues that PMEMD.CUDA is crashing and that this combination is not possible?
Best,
Parker
________________________________________
From: Parker de Waal
Sent: Thursday, September 25, 2014 11:26 PM
To: AMBER Mailing List
Subject: RE: [AMBER] Combined cartesian and NMR restraints
Hi David,
Thank you for the input. I've adjusted the input per your suggestion as follows:
NVT prod
&cntrl
irest=1, ntx =5,
nstlim=5000000, dt=0.002,
ntpr=500, ntwr=500, ntwx=500,
temp0=300.0,
ntt=3, gamma_ln=5, ntb=1, ntp=0,
ntc=2, ntf=2, ig=-1,
cut=9,
ioutfm = 1,
iwrap=1, nmropt=1,
ntr=1, restraint_wt=20., restraintmask=':1-708',
/
&wt type='END' /
DISANG=rst.dist
The only problem now is that when I run pmemd.CUDA the follow error occurs:
Error: an illegal memory access was encountered launching kernel kClearForces
cudaFree GpuBuffer::Deallocate failed an illegal memory access was encountered
I'm currently using the most recent version of AMBER14 (updated and recompiled today) with NVIDIA Driver Version: 340.32 and Cuda compilation tools, release 6.5, V6.5.12 (I'm running arch so rolling back is difficult, but I'm trying now to 5.5)
Parker
________________________________________
From: David A Case [case.biomaps.rutgers.edu]
Sent: Thursday, September 25, 2014 10:56 PM
To: AMBER Mailing List
Subject: Re: [AMBER] Combined cartesian and NMR restraints
On Fri, Sep 26, 2014, Parker de Waal wrote:
>
> After reading through some of the old mailing list threads regarding
> the various workarounds for ntr =1 and nmropt=1, I still need some
> clarification.
>
> I have an NVT http://scanmail.trustwave.com/?c=129&d=49Wk1K_bbQBiQ3IZT3Q0OKQyeRAK9WcrWuJ7VB2iCg&u=http%3a%2f%2fprod%2ein as show below:
>
>
> Now I wish to add cartesian restraints, following the snippet here:
> http://scanmail.trustwave.com/?c=129&d=49Wk1K_bbQBiQ3IZT3Q0OKQyeRAK9WcrWuUuUBb1CQ&u=http%3a%2f%2fambermd%2eorg%2fQuestions%2fconstraints%2ehtml into my rst.dist file as
> seen below:
The group cards for cartesian restraints go into the mdin file, not into the
restraints file.
>
> # N373 D69 VS
> &rst
> ixpk= 0, nxpk= 0, iat= 5235, 6651, r1= 0.00, r2= 0.00, r3= 6, r4= 7,
> rk2=0.0, rk3=25.0, ir6=1, ialtd=0,
> &end
> ... (lines removed)
> # A150 F244 VS
> &rst
> ixpk= 0, nxpk= 0, iat= 2310, 9472, r1= 0.00, r2= 0.00, r3= 6, r4= 7,
> &end
> Hold GPCR fixed
> 20.0
> RES 1 708
> END
> END
>
> However whenever I run this the follow error occurs:
>
> 5. REFERENCE ATOM COORDINATES
>
> default_name
> ----- READING GROUP 1; TITLE:
> DISANG=rst.dist
Note that the program thinks the DISANG card is the title for the cartesian
restraints; so somehow it is not reading the NMR-like restraints correctly.
The first thing you need to do is set ntr=0, and make sure that you are
getting the NMR restraints correctly processed. Follow the example in the
$AMBERHOME/test/nmr directory.
By far the simplest thing to do for cartsian restraints is to set them
in the cntrl namelist:
ntr=1, restraint_wt=20., restraintmask=':1-708',
Then there is no need for any GROUP input.
Beyond this:
You could try using the correct "/" rather than the obsolete (and never
standard "&end" to end namelists. But that is probably not the source of the
problem. You may need to let us know which program you are running
(sander or pmemd?) and which version. Posting the entire mdout file is
probably also required...there is something subtle that is preventing your
file from being parsed correctly.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://scanmail.trustwave.com/?c=129&d=5NWk1DkJ-rwIWc9WjII5xYe7GxkTZDMTXvE-6uITXQ&u=http%3a%2f%2flists%2eambermd%2eorg%2fmailman%2flistinfo%2famber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 26 2014 - 06:00:04 PDT
